CN104599249A - Cable rope bearing bridge deck vehicle load distribution real-time detection method - Google Patents

Abstract

The invention discloses a cable rope bearing bridge deck vehicle load distribution real-time detection method. The cable rope bearing bridge deck vehicle load distribution real-time detection method comprises the steps of shooting a bridge deck image; conducting perspective correction processing and enhancement processing on the image so as to obtain a bridge deck image subjected to image enhancement and adopting an edge information based detection method to judge a bridge deck vehicle; tracking a vehicle image and correcting a deformed structure of the vehicle in the acquired image; using the bridge deck as an absolute coordinate system, accurately drawing a running track of an automobile tire in each video picture section on the bridge deck, splicing tracks of the same vehicle in different video pictures according to a same track principle to obtain a running track of each vehicle on the bridge deck and achieving vehicle load real-time tracking. The cable rope bearing bridge deck vehicle load distribution real-time detection method is wide in application range, low in cost, capable of obtaining a detection result in real time and small in environmental influence. By means of a perspective correction and image enhancement technology, high-quality earlier-stage images can be obtained. The vehicle blocking problem caused by image acquisition in an oblique state can be well solved by means of a dynamic template matching technology, and the method has a wide application range.

Description

Cableway platform bridge floor carload distribution real-time detection method

Technical field

The invention belongs to bridge machinery field, relate to a kind of detection method of cableway platform load, be specifically related to a kind of cableway platform bridge floor vehicle dynamic load assignment detection method.

Background technology

The variable load of highway bridge is mainly vehicular load, is the Main Basis of bridge structure design.Although Bridge Design specification has clear and definite regulation to the distribution form of vehicular load and intensity, but this mobile load arrangement form still brings some problem in Practical Project, still can not meet the analysis and design that becomes more meticulous of bridge structure, be mainly reflected in three aspects.The first, the highway bridge vehicle active load defined in specification is a standardized pattern, and this mode accurately cannot reflect vehicle distribution on the bridge under some extreme case, as traffic congestion on bridge during generation traffic hazard.In this case, vehicular load intensity, likely far away higher than design load, therefore needs to carry out fail-safe analysis by statistical method to extreme case.The second, the mission life of steel bridge deck is limited by the fatigue behaviour of steel, and its life-span also depends on live loading on bridge, the information such as the position need accurately grasp mobile load to act on bridge floor to the assessment of fatigue of steel structures and translational speed.3rd, how easily there is girder and laterally topple in the bridge of some single-columned pier, such accident also occurred under Under Eccentric Loading.When considering that version is arranged, just needing the accurate possibility passed judgment on this extreme event and occur, namely grasping the extreme case of heavy-duty vehicle along outermost lanes.Generally speaking, setting up a rational vehicle active load model has been the task of top priority, and the detection of bridge vehicular load DYNAMIC DISTRIBUTION will be the basis setting up reasonable vehicle active load model.

Bridge vehicular load detection technique has higher practical value in the many aspects such as design, management and maintenance of bridge structure.At present, bridge vehicular load detection technique relates to two aspects.On the one hand, realize for the bridge floor vehicle location detecting image recognition technology that adopts, the mode of large-scale application on bridge, fixes a point to arrange some monitoring cameras be used for grasping bridge floor vehicle condition more.But the limitation of this method is the bridge floor vehicle distribution only can grasped near camera, cannot grasp full-bridge state.In addition, carry out in full-bridge scope the version that vehicle identification is then limited to bridge itself.On the other hand, in order to grasp vehicle weight on bridge, the dynamic weighing mode that adopts realizes more.Although this method can grasp the accurate weight of vehicle, combine if do not distributed with vehicle, also cannot grasp the effect course of load on bridge floor.

Summary of the invention

For deficiency and the defect of existing detection method, the object of the invention is to, a kind of bridge floor vehicle distribution detection method being applicable to cableway platform is provided, and the scope of application of the method is provided according to bridge floor air visibility and spanning size, the method is based on computer vision technique, adopt the methods such as perspective correction, image enhaucament, multilane vehicle detection and Deformable Template, realize the real-time follow-up of vehicular load distribution, for setting up rational mobile load model and Bridge Design provides important evidence.

Step one: shooting bridge floor image; Install dynamic weigher in the porch, each track of cableway platform and obtain vehicle weight information, each bridge tower is installed several video cameras and obtains each section of bridge floor train flow diagram picture, whole visuals field of video camera can cover complete bridge floor; The first, grasp air visibility according to bridge location place meteorological data, simultaneously according to the sensitivity of image capture device, determine the applicability of this method;

The second, for design proportion determination this method of cable-stayed bridge and suspension bridge in the reliability of carrying out vehicle detection and identification.Inclination minimum during suggestion picture photographing is not less than 6 °, too flat cableway platform is applied the method and appearance is gathered more vehicle overlap in picture, easily produce vehicle identification mistake.

Step 2: perspective image correction; The bridge floor image that video camera obtains is carried out perspective correction process, obtains bridge floor front elevation:

(2.1) utilize 4 straight lines (4 straight lines that track surrounds) needed for Hough transform extraction, and identify 4 straight lines that track surrounds;

(2.2) when extracting straight line, record the end points of every bar straight line, extracted the intersection point of the rear result calculated line to Hough transform, the intersection point obtained is exactly the end points of 4 straight lines;

(2.3) using the intersecting point coordinate of 4 straight lines and corresponding actual coordinate, totally 4 pairs of coordinate datas are as the known quantity of separating perspective parameter matrix, and each pixel coordinate in note fault image is designated as (x ₁, y ₁) (x ₂, y ₂) (x ₃, y ₃) (x ₄, y ₄), the coordinate of the point in corresponding front elevation is designated as (u ₁, v ₁) (u ₂, v ₂) (u ₃, v ₃) (u ₄, v ₄), solve 8 perspective parameter vectors by following formula;

$\left[\begin{array}{c}{u}_1\\ v_1\\ u_2\\ v_2\\ u_3\\ v_3\\ u_4\\ v_4\end{array}\right]=\left[\begin{array}{cccccccc}{x}_1& y_1& 1& 0& 0& 0& -u_1{x}_1& -u_1{y}_1\\ 0& 0& 0& x_1& y_1& 1& -v_1{x}_1& -v_1{y}_1\\ x_2& y_2& 1& 0& 0& 0& -u_2{x}_2& -u_2{y}_2\\ 0& 0& 0& x_2& y_2& 1& -v_2{x}_2& -v_2{y}_2\\ x_3& y_3& 1& 0& 0& 0& -u_3{x}_3& -u_3{y}_3\\ 0& 0& 0& x_3& y_3& 1& -v_3{x}_3& -v_3{y}_3\\ x_4& y_4& 1& 0& 0& 0& -u_4{x}_4& -u_4{y}_4\\ 0& 0& 0& x_4& y_4& 1& -v_4{x}_4& -v_4{y}_4\end{array}\right]\×\left[\begin{array}{c}a\\ b\\ c\\ d\\ e\\ f\\ m\\ l\end{array}\right]$

In formula:

(x ₁, y ₁) (x ₂, y ₂) (x ₃, y ₃) (x ₄, y ₄) represent the coordinate of 4 straight-line intersections in distortion figure;

(u ₁, v ₁) (u ₂, v ₂) (u ₃, v ₃) (u ₄, v ₄) represent the coordinate of 4 straight-line intersections in corresponding front elevation;

[a b c d e f m l] ^-18 perspective parameter vectors;

(2.4), after obtaining perspective parameter, the perspective transform adopting the point-to-point mode of image to carry out two-dimensional image calculates, and utilizes following formula to obtain the front elevation of standard;

$\left[\begin{array}{c}u\\ v\end{array}\right]=\left[\begin{array}{cccccccc}x& y& 1& 0& 0& 0& -\mathrm{ux}& -\mathrm{uy}\\ 0& 0& 0& x& y& 1& -\mathrm{vx}& -\mathrm{vy}\end{array}\right]\left[\begin{array}{c}a\\ b\\ c\\ d\\ e\\ f\\ m\\ l\end{array}\right]$

In formula:

[a b c d e f m l] ^-18 perspective parameter vectors;

(x, y) represents the coordinate of distortion figure mid point;

The coordinate that (u, v) puts after representing perspective correction;

Step 3: image enhaucament; Histogram Matching (regulationization) is utilized to carry out image enhaucament to the front elevation after perspective correction process.Concrete formula is as follows:

$s=T\left(r\right)={\∫}_0^r{p}_r\left(x\right)\mathrm{dx}$

$H\left(z\right)={\∫}_0^z{p}_z\left(x\right)\mathrm{dx}=s$

z＝H ^-1(s)＝H ^-1[T(r)]

In formula: r and z represents the gray level of input picture and output image respectively, p _rx () is the probability density function of input gray grade, p _zx () is the probability density function of output gray level;

Step 4: bridge floor vehicle judges; Bridge floor image after obtaining image enhaucament, adopts each section train flow diagram picture of detection method to bridge tower photographs based on marginal information to carry out vehicle detection:

(4.1) Sobel operator is used to carry out extraction edge to each two field picture;

(4.2) carry out threshold process and superposition to the outline map extracted, concrete formula is as follows:

${g^{\′}}_i(x,y)=\left\{\begin{array}{ccc}g& \mathrm{if}& g_i(x,y)>g\\ g_i(x,y)& \mathrm{if}& g_i(x,y)\≤g\end{array}\right.$

In formula: g _iand g' _ibe the outline map before and after threshold process respectively, g is threshold value;

(4.3) superposed by the outline map after threshold process, concrete formula is as follows:

$b_0(x,y)=\underset{i=1}{\overset{25}{\mathrm{\Σ}}}{g^{\′}}_i(x,y)$

(4.4) binary conversion treatment obtains background edge figure, and concrete formula is as follows:

$b(x,y)=\left\{\begin{array}{ccc}1& \mathrm{if}& b_0(x,y)>b\\ 0& \mathrm{if}& b_0(x,y)\≤b\end{array}\right.$

In formula: b is the threshold value of setting;

(4.5) utilize inclusive difference to obtain vehicle movement outline map, concrete formula is as follows:

$i(x,y)=\left\{\begin{array}{ccc}0& \mathrm{if}& b(x,y)=1\\ i(x,y)& \mathrm{if}& b(x,y)=0\end{array}\right.$

In formula: i (x, y) is movement edge figure;

(4.6) arrange a band strip detection zone at image near bridge tower position, and be divided into one group of continuous print window, concrete formula is as follows:

$h\left[i\right]=\underset{y=i\×w}{\overset{i\×w+w}{\mathrm{\Σ}}}\underset{x=0}{\overset{m}{\mathrm{\Σ}}}i(x,y)$

In formula: m is the height of window, w is the width of window;

(4.7) whether what threshold method judged each window of movement edge figure in detection zone is effective information, and obtains one group 0,1 binary sequence l [i], and concrete formula is as follows:

$l\left[i\right]=\left\{\begin{array}{ccc}1& \mathrm{if}& h\left[i\right]>h\\ 0& \mathrm{if}& h\left[i\right]\≤h\end{array}\right.$

(4.8) add up to the marginal information in vehicle passing detection district, thus preserve complete information of vehicles (l [i] continuous renewal), concrete formula is as follows:

$f\left[i\right]=\left\{\begin{array}{ccc}1& \mathrm{if}& l\left[i\right]=1\\ f\left[i\right]& \mathrm{if}& l\left[i\right]=0\end{array}\right.$

In formula: f [i] is the information of vehicles after adding up;

(4.9) when continuous print some 1 being detected in array f [i], when being some continuous print 0 in the array l [i] of correspondence, being then determined with car and passing through;

Step 5: vehicle image is followed the trail of; After determining vehicle passing detection band, Deformable Template technology is adopted to follow the trail of the wheelpath of each automobile:

(5.1) after determining vehicle passing detection band, in former figure and outline map, vehicle ' s contour angular region near detection zone, is chosen respectively as template according to vehicle width and the speed of a motor vehicle, generate edge image template, edge image template carries out threshold process, obtains left hand edge (or right hand edge) position a and the upper marginal position b of vehicle ' s contour;

(5.2) note obtains the image of template is K two field picture (t), in K-1 two field picture (t+ Δ t), use related function to calculate, carry out template matches identification, matching range is the region that original template the right and left and top expand a vehicle width;

(5.3) matching area is obtained according to related function peak value, matching area edge is identified, obtain the position a' of tailstock left hand edge (or right hand edge) and the position b' of coboundary, compare with original value a, b, make and revise adjustment, and revised region is designated as new template;

(5.4) repeat (5.2) and (5.3) to information of vehicles real-time update, obtain the driving trace of automobile revolver (or right take turns) in entire picture, and be plotted on whole bridge floor;

Step 6: malformation correction; In shooting process, Vehicle Load will cause bridge structure to deform, and slight corner or lateral shift appear in bridge tower top, cause vehicle gathering the relative movement in picture, need to revise this part additional displacement;

(6.1) before carrying out the collection of bridge floor vehicle picture, in both sides, bridge floor track by gauged distance placement of images gauge point or artificial cognition to the position with notable feature, by catching gauge point to after the image recognition collected, obtain girder bulk deformation feature;

(6.2) according to the main beam deformation feature obtained, additional displacement correction is carried out to the vehicle location captured;

Step 7: after obtaining the doughnut driving trace of each section of video, take bridge floor as absolute coordinate system, bridge floor draws the driving trace of doughnut in every section of video pictures accurately, according to track same principle, the track of same vehicle in different video picture is spliced, obtain the driving trace of each vehicle at bridge floor, finally realize the real-time follow-up of cableway platform bridge floor vehicular load.According to the distance between the pull-in time of each point in track of vehicle and, calculate the rate curve of vehicle by bridge floor at adjacent 2.

Compare existing method, the present invention has following beneficial outcomes:

This method is applied widely, with low cost, and equipment needed thereby is installed simple and easy, can Real-time Obtaining testing result, affected by environment little; Use perspective correction and image enhancement technique can obtain the image in early stage of better quality; By bridge floor gauge point mode, the vehicle location additional displacement that malformation causes is revised, obtain more accurate recognition result; Use arranges image detection zone, accurately identifies the many vehicles of multilane; Use Deformable Template technology can solve the occlusion problem that the collection of heeling condition hypograph causes preferably, makes this invention have applicability widely.

Accompanying drawing explanation

Fig. 1 is cableway platform bridge floor load assignment real-time tracking system structural drawing.

Fig. 2 is cableway platform bridge floor load assignment real-time follow-up overhaul flow chart.

Fig. 3 is the vehicle checking method process flow diagram based on marginal information.

Fig. 4 is Deformable Template process flow diagram.

Fig. 5 is perspective transform principle schematic.

Fig. 6 is a certain frame bridge floor image of video camera shooting.

Fig. 7 is the image after Fig. 6 carries out perspective correction.

Fig. 8 is the edge image of Fig. 7 after Sobel algorithm.

Fig. 9 is background edge image.

Figure 10 is the movement edge hum pattern that Fig. 8 and background edge figure do pardon difference gained.

Figure 11 is that bridge floor road map is as detection zone.

Figure 12 is that a certain car is blocking primary template figure and the outline map thereof of Deformable Template in special case.

Figure 13 is that a certain car is blocking the search matching area of certain frame template in special case.

Figure 14 is that a certain car is blocking in special case the partial target image realizing Deformable Template.

Figure 15 is the driving trace figure blocking special case.

Embodiment

Cableway platform bridge floor vehicle dynamic load assignment detection method of the present invention, as shown in Fig. 1 to Figure 15, dynamic weigher is installed in the porch, each track of cableway platform and obtains vehicle weight information, bridge tower is installed video camera and obtains bridge floor train flow diagram picture, the coverage of video camera can cover whole bridge floor; The bridge floor image that video camera obtains is carried out perspective correction process, obtains the image of normal vehicle size; After the image obtaining perspective correction, histogram matching (histogram equalization method) is adopted to strengthen image; Whether adopt the method based on marginal information to detect has moving target to pass through; The basis passed through there being vehicle adopt the method for Deformable Template follow the trail of the driving trace of vehicle; By the seizure identification of bridge floor special sign thing, vehicle coordinate is corrected, offset the additional displacement because malformation produces; Adopt track same principle that the track of same vehicle in different video picture is spliced, obtain the driving trace of each vehicle at bridge floor, finally realize the real-time follow-up of cableway platform bridge floor vehicular load.

Cableway platform bridge floor vehicular load distribution real-time follow-up calculates vehicular load mainly through the weight of the distribution of vehicle on bridge floor and vehicle and speed.Therefore, obtain the position of vehicle, speed, weight is the key detected.Along with developing rapidly of Digital image technology, the vehicle testing techniques based on video reaches its maturity, and as prior art, can carry out respective handling, meet needs of the present invention.These technology all have low cost, high accuracy, high efficiency feature.

Long span stayed-cable bridge for certain the two rope face of employing, H type bridge tower carries out application of the present invention.This bridge main span be 450m, H type bridge tower thwart beam end face to the vertical distance of bridge floor be 56m.According to on-the-spot physical condition, be selected in crossbeam end face and image collecting device is set.Air visibility when carrying out image acquisition is more than 1000m.Single camera head covers half main span or full end bay, and therefore minimum sight line inclination angle is 14 °, therefore meets use pacing items of the present invention.

Fig. 1 is cableway platform bridge floor vehicular load distribution real-time tracking system structural drawing, for double tower carrying bogie, in porch, bridge track, dynamic weigher is installed, two video cameras arranged by each bridge tower, regulates camera installation locations to enable these video camera coverages cover whole bridge floor.Obtain vehicle weight information and velocity information, dynamic weigher energy direct convenience ground obtains vehicle weight information; And vehicle speed information, then according to the distance between the pull-in time of each point in track of vehicle and, calculate the rate curve of vehicle by bridge floor at adjacent 2.

Fig. 2 is cableway platform bridge floor vehicular load distribution real-time tracking system process flow diagram, and first, bridge floor wagon flow video is beamed back background computer by the video camera that bridge tower top is installed, and processes video image successively in order.The image of Hough transform and perspective transform is adopted to carry out perspective correction; After corrected perspective image, utilize the gauge point on bridge floor to carry out malformation correction, this example directly adopts the expansion joint on bridge floor to carry out position correction; Then histogram equalization is adopted to carry out enhancing process to the image after correction; After image pre-treatment, adopt the method based on marginal information to arrange a detection again near the position of bridge tower and bring and judged whether that vehicle passes through, detailed process as shown in Figure 3; After confirmation has vehicle passing detection district, adopt the method for Deformable Template to follow the trail of vehicle driving trace, detailed process as shown in Figure 4; Finally adopt track same principle that the track of same vehicle in different video picture is spliced, obtain the driving trace of each vehicle at bridge floor, finally realize the real-time follow-up of cableway platform bridge floor vehicular load.Embodiment:

Step one: install dynamic weigher in the porch, each track of cableway platform and obtain vehicle weight, the video camera installing or more at two bridge tower tops obtains bridge floor wagon flow image/video, and the visual field of video camera can cover whole bridge floor.

Step 2: the bridge floor image that video camera obtains is carried out perspective correction process, obtains bridge floor front elevation, perspective correction schematic diagram as shown in Figure 5:

(2.1) utilize Hough transform extract needed for 4 straight lines (4 straight lines that track surrounds, as shown in Figure 6);

(2.2) when extracting straight line, record the end points of every bar straight line, extracted the intersection point of the rear result calculated line to Hough transform, the intersection point obtained is exactly the end points of 4 straight lines;

(2.3) using the intersecting point coordinate of 4 straight lines and corresponding actual coordinate totally 4 pairs of coordinate datas as the known quantity of separating perspective parameter matrix, solve perspective parameter vector by following formula:

$\left[\begin{array}{c}{u}_1\\ v_1\\ u_2\\ v_2\\ u_3\\ v_3\\ u_4\\ v_4\end{array}\right]=\left[\begin{array}{cccccccc}{x}_1& y_1& 1& 0& 0& 0& -u_1{x}_1& -u_1{y}_1\\ 0& 0& 0& x_1& y_1& 1& -v_1{x}_1& -v_1{y}_1\\ x_2& y_2& 1& 0& 0& 0& -u_2{x}_2& -u_2{y}_2\\ 0& 0& 0& x_2& y_2& 1& -v_2{x}_2& -v_2{y}_2\\ x_3& y_3& 1& 0& 0& 0& -u_3{x}_3& -u_3{y}_3\\ 0& 0& 0& x_3& y_3& 1& -v_3{x}_3& -v_3{y}_3\\ x_4& y_4& 1& 0& 0& 0& -u_4{x}_4& -u_4{y}_4\\ 0& 0& 0& x_4& y_4& 1& -v_4{x}_4& -v_4{y}_4\end{array}\right]\×\left[\begin{array}{c}a\\ b\\ c\\ d\\ e\\ f\\ m\\ l\end{array}\right]$

In formula:

(x ₁, y ₁) (x ₂, y ₂) (x ₃, y ₃) (x ₄, y ₄) represent the coordinate of 4 straight-line intersections in distortion figure;

(u ₁, v ₁) (u ₂, v ₂) (u ₃, v ₃) (u ₄, v ₄) represent the coordinate of 4 straight-line intersections in corresponding front elevation;

[a b c d e f m l] ^-18 perspective parameter vectors;

(2.4), after obtaining perspective parameter, the perspective transform adopting the point-to-point mode of image to carry out two-dimensional image calculates, and utilizes following formula to obtain the front elevation of standard;

$\left[\begin{array}{c}u\\ v\end{array}\right]=\left[\begin{array}{cccccccc}x& y& 1& 0& 0& 0& -\mathrm{ux}& -\mathrm{uy}\\ 0& 0& 0& x& y& 1& -\mathrm{vx}& -\mathrm{vy}\end{array}\right]\left[\begin{array}{c}a\\ b\\ c\\ d\\ e\\ f\\ m\\ l\end{array}\right]$

In formula:

[a b c d e f m l] ^-18 perspective parameter vectors;

(x, y) represents the coordinate of distortion figure mid point;

The coordinate that (u, v) puts after representing perspective correction;

Correct result as shown in Figure 7;

Step 3: utilize Histogram Matching (regulationization) to carry out image enhaucament to the front elevation after perspective correction and rotation correction process, concrete formula is as follows:

$s=T\left(r\right)={\∫}_0^r{p}_r\left(x\right)\mathrm{dx}$

$H\left(z\right)={\∫}_0^z{p}_z\left(x\right)\mathrm{dx}=s$

z＝H ^-1(s)＝H ^-1[T(r)]

In formula:

R and z represents the gray level of input picture and output image respectively;

P _rx () is the probability density function of input gray grade;

P _zx () is the probability density function of output gray level;

Step 4: the bridge floor image after obtaining image enhaucament, adopts each section train flow diagram picture of detection method to bridge tower photographs based on marginal information to carry out vehicle detection:

(4.1) Sobel operator extraction edge is carried out to each two field picture, as shown in Figure 8;

(4.2) carry out threshold process to all edge images, concrete formula is as follows:

${g^{\′}}_i(x,y)=\left\{\begin{array}{ccc}g& \mathrm{if}& g_i(x,y)>g\\ g_i(x,y)& \mathrm{if}& g_i(x,y)\≤g\end{array}\right.$

In formula:

G _iand g' _ithe outline map before and after threshold process respectively;

G is threshold value, now n=30, g=8;

(4.3) superposed by the outline map after threshold process, concrete formula is as follows:

$b_0(x,y)=\underset{i=1}{\overset{25}{\mathrm{\Σ}}}{g^{\′}}_i(x,y)$

(4.4) binary conversion treatment obtains background edge figure, and concrete formula is as follows:

$b(x,y)=\left\{\begin{array}{ccc}1& \mathrm{if}& b_0(x,y)>b\\ 0& \mathrm{if}& b_0(x,y)\≤b\end{array}\right.$

In formula:

B is the threshold value of setting, and b=220, background edge figure are as shown in Figure 9;

(4.5) utilize inclusive difference to obtain vehicle movement outline map, concrete formula is as follows:

$i(x,y)=\left\{\begin{array}{ccc}0& \mathrm{if}& b(x,y)=1\\ i(x,y)& \mathrm{if}& b(x,y)=0\end{array}\right.$

In formula:

I (x, y) is movement edge figure, as shown in Figure 10;

(4.6) to arrange a long and narrow detection zone near bridge tower position at image, and be divided into one group of continuous print window, concrete formula is as follows:

$h\left[i\right]=\underset{y=i\×w}{\overset{i\×w+w}{\mathrm{\Σ}}}\underset{x=0}{\overset{m}{\mathrm{\Σ}}}i(x,y)$

In formula:

M is the height of window, and m=10, w are the width of window, w=10, and detection zone is as shown in Figure 11 red area;

(4.7) whether what threshold method judged each window of movement edge figure in detection zone is effective information, and obtains one group 0,1 binary sequence l [i], and concrete formula is as follows:

$l\left[i\right]=\left\{\begin{array}{ccc}1& \mathrm{if}& h\left[i\right]>h\\ 0& \mathrm{if}& h\left[i\right]\≤h\end{array}\right.$

(4.8) add up to the marginal information in vehicle passing detection district, thus preserve complete information of vehicles (l [i] continuous renewal), concrete formula is as follows:

$f\left[i\right]=\left\{\begin{array}{ccc}1& \mathrm{if}& l\left[i\right]=1\\ f\left[i\right]& \mathrm{if}& l\left[i\right]=0\end{array}\right.$

In formula:

F [i] is the information of vehicles after adding up;

(4.9) when continuous print some 1 being detected in array f [i], when being some continuous print 0 in the array l [i] of correspondence, being then determined with car and passing through;

Step 5: after defining vehicle passing detection band, adopts Deformable Template technology to follow the trail of the wheelpath of each automobile:

(5.1) after determining vehicle passing detection band, note frame number is now K, makes K=K-LINE, and LINE is the frame number (in stream video of picking up the car in the present embodiment, the vehicle that is blocked is demonstrated as special case, LINE=12) of vehicle passing detection band;

(5.2) in K two field picture, in former figure and outline map, vehicle tail near detection zone, is automatically chosen respectively as template and border template according to vehicle width and the speed of a motor vehicle, as shown in figure 12;

(5.3) edge template carries out threshold process, obtains the position a of tailstock left hand edge (or right hand edge) and the position b of coboundary;

(5.4) note obtains the image of template is K two field picture, in K-1 two field picture, use related function to calculate, carry out template matches identification, matching range is the region that original template surrounding expands a vehicle width, vehicle width is 8, and matching range as shown in figure 13;

(5.5) matching area is obtained according to related function peak value, matching area edge is identified, obtain the position a ' of tailstock left hand edge (or right hand edge) and the position b ' of coboundary, compare with original value a, b, make and revise adjustment, and revised region is designated as new template, template place image is designated as K two field picture;

(5.6) repeat the operation of aforementioned (5.2) to (5.5), to information of vehicles real-time update, obtain the driving trace of automobile revolver (or right take turns) in entire picture, and be plotted on whole bridge floor;

It is the partial frame number template matching results of stagnant Deformable Template after blocking special case shown in Figure 14;

The driving trace figure blocking special case shown in Figure 15;

Step 6: in shooting process, produce vehicle has certain additional displacement because vehicular load causes malformation in collection picture, need choose the gauge point that bridge floor is arranged and carries out physical location seizure and correct vehicle location.The present embodiment directly catches the deck expansion joint position in picture, obtains the position correction value of each car in each moment.Namely the present invention is by choosing deck expansion joint as gauge point, carry out physical location seizure and correct vehicle location, obtain the position correction value of each car in each moment, just can eliminate vehicular load cause malformation and the vehicle that produces at the additional displacement gathering picture.

Step 7: after obtaining the doughnut driving trace of each section of video, take bridge floor as absolute coordinate system, bridge floor draws the driving trace of doughnut in every section of video pictures accurately, according to track same principle, the track of same vehicle in different video picture is spliced, obtain the driving trace of each vehicle at bridge floor, finally realize the real-time follow-up of cableway platform bridge floor vehicular load.

What finally illustrate is, above embodiment is only in order to illustrate technical scheme of the present invention and unrestricted, although with reference to preferred embodiment to invention has been detailed description, those of ordinary skill in the art is to be understood that, can modify to technical scheme of the present invention or equivalent replacement, and not departing from aim and the scope of technical solution of the present invention, it all should be encompassed in the middle of right of the present invention.

Claims (2)

1. cableway platform bridge floor carload distribution real-time detection method, is characterized in that comprising the following steps:

One, bridge floor image is taken; Install dynamic weigher in the porch, each track of cableway platform and obtain vehicle weight information, each bridge tower is installed several video cameras and obtains each section of bridge floor train flow diagram picture, whole visuals field of video camera can cover complete bridge floor; Obtain vehicle speed information;

Two, perspective image correction; The bridge floor image that video camera obtains is carried out perspective correction process, obtains bridge floor front elevation:

(2.1) utilize Hough transform to extract 4 straight lines surrounded needed for track, and identify 4 straight lines that track surrounds;

(2.2) when extracting straight line, record the end points of every bar straight line, extracted the intersection point of the rear result calculated line to Hough transform, the intersection point obtained is exactly the end points of 4 straight lines;

(2.3) using the intersecting point coordinate of 4 straight lines and corresponding actual coordinate, totally 4 pairs of coordinate datas are as the known quantity of separating perspective parameter matrix, and each pixel coordinate in note fault image is designated as (x ₁, y ₁) (x ₂, y ₂) (x ₃, y ₃) (x ₄, y ₄), the coordinate of the point in corresponding front elevation is designated as (u ₁, v ₁) (u ₂, v ₂) (u ₃, v ₃) (u ₄, v ₄), solve 8 perspective parameter vectors by following formula;

In formula: $\left[\begin{array}{c}{u}_1\\ v_1\\ u_2\\ v_2\\ u_3\\ v_3\\ u_4\\ v_4\end{array}\right]=\left[\begin{array}{cccccccc}{x}_1& y_1& 1& 0& 0& 0& -u_1{x}_1& -u_1{y}_1\\ 0& 0& 0& x_1& y_1& 1& -v_1{x}_1& -v_1{y}_1\\ x_2& y_2& 1& 0& 0& 0& -u_2{x}_2& -u_2{y}_2\\ 0& 0& 0& x_2& y_2& 1& -v_2{x}_2& -v_2{y}_2\\ x_3& y_3& 1& 0& 0& 0& -u_3{x}_3& -u_3{y}_3\\ 0& 0& 0& x_3& y_3& 1& -v_3{x}_3& -v_3{y}_3\\ x_4& y_4& 1& 0& 0& 0& -u_4{x}_4& -u_4{y}_4\\ 0& 0& 0& x_4& y_4& 1& -v_4{x}_4& -v_4{y}_4\end{array}\right]\×\left[\begin{array}{c}a\\ b\\ c\\ d\\ e\\ f\\ m\\ l\end{array}\right]$

(x ₁, y ₁) (x ₂, y ₂) (x ₃, y ₃) (x ₄, y ₄) represent the coordinate of 4 straight-line intersections in distortion figure;

(u ₁, v ₁) (u ₂, v ₂) (u ₃, v ₃) (u ₄, v ₄) represent the coordinate of 4 straight-line intersections in corresponding front elevation;

[a b c d e f m l] ^-18 perspective parameter vectors;

(2.4), after obtaining perspective parameter, the perspective transform adopting the point-to-point mode of image to carry out two-dimensional image calculates, and utilizes following formula to obtain the front elevation of standard;

$\left[\begin{array}{c}u\\ v\end{array}\right]=\left[\begin{array}{cccccccc}x& y& 1& 0& 0& 0& -\mathrm{ux}& -\mathrm{uy}\\ 0& 0& 0& x& y& 1& -\mathrm{vx}& -\mathrm{vy}\end{array}\right]\left[\begin{array}{c}a\\ b\\ c\\ d\\ e\\ f\\ m\\ l\end{array}\right]$

In formula:

[a b c d e f m l] ^-18 perspective parameter vectors;

(x, y) represents the coordinate of distortion figure mid point;

The coordinate that (u, v) puts after representing perspective correction;

Three, image enhaucament; Histogram Matching (regulationization) is utilized to carry out image enhaucament to the front elevation after perspective correction process; Concrete formula is as follows:

$s=T\left(r\right)={\∫}_0^r{p}_r\left(x\right)\mathrm{dx}$

$H\left(z\right)={\∫}_0^z{p}_z\left(x\right)\mathrm{dx}=s$

z＝H ^-1(s)＝H ^-1[T(r)]

In formula: r and z represents the gray level of input picture and output image respectively, p _rx () is the probability density function of input gray grade, p _zx () is the probability density function of output gray level;

Four, bridge floor vehicle judges; Bridge floor image after obtaining image enhaucament, adopts each section train flow diagram picture of detection method to bridge tower photographs based on marginal information to carry out vehicle detection:

(4.1) Sobel operator is used to carry out extraction edge to each two field picture;

(4.2) carry out threshold process and superposition to the outline map extracted, concrete formula is as follows:

${g^{\′}}_i(x,y)=\left\{\begin{array}{ccc}g& \mathrm{if}& g_i(x,y)>g\\ g_i(x,y)& \mathrm{if}& g_i(x,y)\≤g\end{array}\right.$

In formula: g _iwith g ' _ibe the outline map before and after threshold process respectively, g is threshold value;

(4.3) superposed by the outline map after threshold process, concrete formula is as follows:

$b_0(x,y)=\underset{i=1}{\overset{25}{\mathrm{\Σ}}}{g^{\′}}_i(x,y)$

(4.4) binary conversion treatment obtains background edge figure, and concrete formula is as follows:

$b(x,y)=\left\{\begin{array}{ccc}1& \mathrm{if}& b_0(x,y)>b\\ 0& \mathrm{if}& b_0(x,y)\≤b\end{array}\right.$

In formula: b is the threshold value of setting;

(4.5) utilize inclusive difference to obtain vehicle movement outline map, concrete formula is as follows:

$i(x,y)=\left\{\begin{array}{ccc}0& \mathrm{if}& b(x,y)=1\\ i(x,y)& \mathrm{if}& b(x,y)=0\end{array}\right.$

In formula: i (x, y) is movement edge figure;

(4.6) arrange a band strip detection zone at image near bridge tower position, and be divided into one group of continuous print window, concrete formula is as follows:

$h\left[i\right]=\underset{y=i\×w}{\overset{i\×w+w}{\mathrm{\Σ}}}\underset{x=0}{\overset{m}{\mathrm{\Σ}}}i(x,y)$

In formula: m is the height of window, w is the width of window;

(4.7) whether what threshold method judged each window of movement edge figure in detection zone is effective information, and obtains one group 0,1 binary sequence l [i], and concrete formula is as follows:

$l\left[i\right]=\left\{\begin{array}{ccc}1& \mathrm{if}& h\left[i\right]>h\\ 0& \mathrm{if}& h\left[i\right]\≤h\end{array}\right.$

(4.8) add up to the marginal information in vehicle passing detection district, thus preserve complete information of vehicles (l [i] continuous renewal), concrete formula is as follows:

$f\left[i\right]=\left\{\begin{array}{ccc}1& \mathrm{if}& l\left[i\right]=1\\ f\left[i\right]& \mathrm{if}& l\left[i\right]=0\end{array}\right.$

In formula: f [i] is the information of vehicles after adding up;

(4.9) when continuous print some 1 being detected in array f [i], when being some continuous print 0 in the array l [i] of correspondence, being then determined with car and passing through;

Five, vehicle image is followed the trail of; After determining vehicle passing detection district, Deformable Template technology is adopted to follow the trail of the wheelpath of each automobile:

(5.1) after determining vehicle passing detection band, in former figure and outline map, vehicle ' s contour angular region near detection zone, is chosen respectively as template according to vehicle width and the speed of a motor vehicle, generate edge image template, edge image template carries out threshold process, obtains left hand edge (or right hand edge) position a and the upper marginal position b of vehicle ' s contour;

(5.2) note obtains the image of template is K two field picture (t), in K-1 two field picture (t+ Δ t), use related function to calculate, carry out template matches identification, matching range is the region that original template the right and left and top expand a vehicle width; The i.e. K two field picture of t, uses related function to calculate with the K-1 two field picture of t+ Δ t, carries out template matches identification;

(5.3) matching area is obtained according to related function peak value, matching area edge is identified, obtain the position a' of tailstock left hand edge (or right hand edge) and the position b' of coboundary, compare with original value a, b, make and revise adjustment, and revised region is designated as new template;

(5.4) repeat (5.2) and (5.3) to information of vehicles real-time update, obtain the driving trace of automobile revolver (or right take turns) in entire picture, and be plotted on whole bridge floor;

Six, malformation correction; In shooting process, Vehicle Load will cause bridge structure to deform, and completely vertical or torsional deflection occur girder, and bridge tower ejects now slight corner, length travel or lateral shift, causing vehicle gathering the relative movement in picture, needing to revise this part additional displacement;

(6.1) before carrying out the collection of bridge floor vehicle picture, in both sides, bridge floor track by gauged distance placement of images gauge point or artificial cognition to the position with notable feature, by catching gauge point to after the image recognition collected, obtaining girder and gathering the bulk deformation feature in image;

(6.2) according to the main beam deformation feature obtained, additional displacement correction is carried out to the vehicle location captured;

Seven, after obtaining the doughnut driving trace of each section of video, take bridge floor as absolute coordinate system, bridge floor draws the driving trace of doughnut in every section of video pictures accurately, according to track same principle, the track of same vehicle in different video picture is spliced, obtain the driving trace of each vehicle at bridge floor, finally realize the real-time follow-up of cableway platform bridge floor vehicular load.

2. according to claim 1 cableway platform bridge floor carload distribution real-time detection method, it is characterized in that: choose deck expansion joint as gauge point, carry out physical location seizure and correct vehicle location, obtain the position correction value of each car in each moment, the vehicle that elimination vehicular load causes malformation and produces is at the additional displacement gathering picture.